Retrieval-Augmented Generation (RAG)
Also called: RAG
Retrieval-augmented generation is an architectural pattern that connects a generative model to a specific body of content, typically proprietary company documents, rather than relying on the model's public training alone. The model retrieves the relevant documents at query time and uses them to ground its response. When an HR chatbot answers a question about your own parental leave policy instead of giving a generic approximation, that is RAG.
In practice
RAG is the cheapest route to a model that knows your business, because it needs no retraining: you change the documents, not the model. The work and the cost therefore sit in the content, so establish whose job it is to keep the source corpus current, and what the system does when two documents contradict each other.